Oldcircle/geo-sleuth ? reverse-engineered prompt

Build me a photo location detective tool that I can hand a single image to and get back where it was taken, plus the direction the camera was facing and how confident it is. It should work even when there is no text or obvious landmark in the photo, by using map geometry, elevation, satellite imagery, and street view style evidence. I want the result to show its work clearly, with the key clues it used, the files or outputs it relied on, and an evidence image that makes the match easy to verify.

Make it usable as an agent skill so it can run from tools like Claude Code, Cursor, Codex, Gemini CLI, OpenCode, and GitHub Copilot. Include simple setup and a quick health check I can run first, then let me give it a photo and ask, find where this photo was taken. If you need current docs or APIs, look them up online first.